Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond
Learning with Kernels: Support Vector Machines, Regularization, Optimization, and Beyond
Pattern Classification (2nd Edition)
Pattern Classification (2nd Edition)
An introduction to kernel-based learning algorithms
IEEE Transactions on Neural Networks
A subject transfer framework for EEG classification
Neurocomputing
Hi-index | 0.01 |
Common spatial patterns (CSP) has proved to be very successful in EEG feature extraction. To relax the presumption of strictly linear patterns in the CSP, nonlinear variants of the approach are proposed using the kernel method. However, they typically suffer from two main drawbacks: the problem of complexity and low generalization ability dealing with different subjects. To overcome these drawbacks, in this paper, two effective solutions are proposed. First, data bunching in the low-dimensional space is used to solve the complexity problem. The second problem is tackled by choosing appropriate kernel functions, which take into account very small amounts of nonlinearity in a generally linear context of the brain spatial patterns, and also are able to be adapted to fit each certain case.